--predict_cv_to_feats

Switch

--predict_cv_to_feats NAME

Description

Runs N-fold cross-validated regression prediction and writes the out-of-fold predictions into a new feature table, without needing a pre-saved model. Aliased as --predict_combo_to_feats, --predict_regression_all_to_feats.

Argument and Default Value

A name used to build the destination table (feat$p_<model>_<name>$<corptable>$<correl_field>). Default: None.

Details

Verified table naming and most of the pipeline work correctly — table creation, N-fold splitting, and prediction all ran ([Inserting Predictions as Feature values for age]).

Known issue — broken in SQLite mode. The final insert calls mm.executeWriteMany(...) with SQL built using MySQL's %s parameter placeholders. SQLite's sqlite3 driver expects ? placeholders instead, so writeCursor.executemany(sql, rows) fails with sqlite3.OperationalError: near "%": syntax error — different root cause from most other SQLite issues found in this pass (a placeholder-style mismatch, not a MySQL-only SQL statement). Works against MySQL.

Other Switches

Required Switches:

• -d, -g, -t

• -f

• --outcome_table, --outcomes

Optional Switches:

• --folds

Example Commands

Requires MySQL — broken against SQLite, see Known issue above:

dlatkInterface.py -d dla_tutorial -t msgs -c user_id -f 'feat$1to3gram$msgs$user_id$0_05' \
--outcome_table blog_outcomes --outcomes age --folds 10 --predict_cv_to_feats cv_age